Surgical necrotizing enterocolitis but not spontaneous intestinal perforation is associated with adverse neurological outcome at school age

Gastrointestinal complications during the neonatal period, i.e. necrotizing enterocolitis (NEC) and spontaneous intestinal perforation (SIP), are associated with adverse short-term outcome in very-low-birthweight infants (VLBWI, <1500 g birth weight). However, little is known about the neurological outcome of survivors at school age. We analysed data of 2241 infants followed-up at the age of 6 years. To determine the effect of NEC and SIP on cognitive outcome in consideration of other important confounding factors, we used multivariable logistic regression models. In addition, infants with surgical diagnosis of NEC (n = 43) or SIP (n = 41) were compared to NEC (n = 43) or SIP (n = 41) negative controls using Mahalanobis distance matching. Infants with a history for NEC had a three times increased risk (RR 3.0 [1.8–4.2], p < 0.001) to develop IQ scores <85 while history of surgical SIP did not increase the relative risk for lower IQs at school age (RR 1.0 [0.4–2.1], p = 1.000). In a matched-cohort analysis, we confirmed that infants with surgical NEC had lower mean IQ results than unaffected controls (±SD) (85±17 vs. 94±14, p = 0.023) while no differences were found for history of SIP. Our results reflect that the different aetiology and inflammatory extent of NEC and SIP may lead to disparate neurodevelopment trajectories. Hence, our data suggest a potential role of early gut-brain axis distortion in infants with NEC which needs to be further explored.


Discussion
In a large cohort of VLBWI from the German Neonatal Network, we found that history of NEC but not SIP is associated with an increased risk of impaired neurological development, in particular reduced IQ scores at 6 years of age. Infants with NEC had a three times increased relative risk to develop IQ results <85. So far, most studies on the neurodevelopment of VLBWI reported outcome at the age of 18-36 months 17 or used data of cohorts from the 1990s 18 which are not necessarily comparable with recent cohorts who have a decreased short-term morbidity including NEC [19][20][21][22] . In one study, the authors found a reduced IQ score in children with a history of gastrointestinal disease (NEC or SIP) 11 . In our study, we made a clear distinction between surgical NEC and surgical SIP based on pathophysiological and macroscopical (as defined by attending surgeon) criteria. The pathophysiology of NEC is based on the immaturity of the gastrointestinal tract and understood as a multifactorial interplay leading to inflammatory processes 23 . Distinct characteristics of bacterial colonization and inappropriate colonization of the premature intestine predispose infants to NEC suggesting a causal relationship between gut bacteria and NEC 24,25 . Together with an inadequate anti-inflammatory response observed in the immature intestine, dysbiosis triggers an inflammatory cascade leading to NEC. Intestinal permeability is observed in NEC and modulated through www.nature.com/scientificreports www.nature.com/scientificreports/ the expression of Toll-like receptors (TLR4) 12 . Nino et al. found a link between NEC and brain injury through activation of TLR4 on microglial cells in the brain. TLR4 stimulation by gut-derived mediators impacts brain injury since TLR4-deficient mice were protected from NEC-induced brain injury. In the setting of prematurity, NEC is associated with a "proinflammatory regulatory protein profile" 15 which suggests a link between sustained inflammation and adverse neurodevelopmental outcomes 2,14 .
In our matched cohort, we found decreased head circumferences in the NEC and SIP group. Some studies report substantial growth failure (<10 th percentile) for weight, length, and head circumference in infants with NEC 26 , but results were conflicting 11 . One study examining head biometrics in magnet resonance imaging (MRI) Figure 1. Enrollment, in-and exclusion for analysis of neurologic and motor development at the age of 6 years. In-and exclusion for motor function and intelligence testing in 6 year old children born as VLBWI; *Reasons for not participating the follow-up assessment despite selection were: no current contact data available n = 1349, parents declined invitation for follow-up n = 519, parents were interested to participate but were not available at suggested follow-up dates n = 131, no-show despite arranged follow-up assessment n = 65; **Reasons for no WPPSI assessment included: WPPSI or other cognitive test within 12 months n = 161, language problems n = 24, child not motivated n = 39, serious disorder not related to prematurity (e.g. trisomy 21) n = 7, other reasons n = 10. (2020) 10:2373 | https://doi.org/10.1038/s41598-020-58761-6 www.nature.com/scientificreports www.nature.com/scientificreports/ after NEC found associations with reduced biparietal width 27 . Another study comparing white matter abnormalities (WMA) on brain MRI in NEC and SIP infants showed that infants with NEC had higher WMA scores than those with SIP 28 assuming a higher vulnerability of oligodendroglial precursors after systemic inflammatory  www.nature.com/scientificreports www.nature.com/scientificreports/ processes such as NEC. In our study, we cannot exactly give an explanation for our findings and can only speculate about causes. It is unclear if our findings are the result of sustained inflammation originating in the neonatal period, the result of chronic malnutrition in this vulnerable population or statistical biased observations. For example, the reduced head circumference in the SIP cohort could be explained by significant reduced birth weight and reduced lower gestational age in the matched SIP group leading to a reduced head growth. We therefore think that our findings concerning growth parameters should be interpreted with caution and further scientific efforts should clarify the role of inflammation or malnutrition in these infants, as interventions here could improve neurologic outcome.
NEC and SIP are both entities with low incidences in a very unique patient cohort. Here, our study has a powerful setting as a large multi-centre study with prospective collection of data and uniform follow-up assessment by the same study team. However, our study also has limitations. We decided to use stringent criteria for NEC (NEC requiring surgery) and SIP (requiring surgery) as primary outcome measure. Hence, the diagnosis of "medical NEC", i.e. NEC not requiring surgery, which might be difficult to distinguish from other entities, was not considered. Second, our NEC or SIP definition is based on the clinical evaluation of the attending surgeon and neonatologist and not necessarily based on histopathology. Third, our follow-up cohort has a risk of selection bias. For the follow-up, we chose a random invitation practice, but were not able to reduce some differences in both groups. Some factors that were associated with better neurological outcomes were more common in children who were followed-up compared to those who were not including higher birth weight and higher exposure rate with antenatal steroids. On the other hand, BPD and multiple birth were overrepresented in the followed-up cohort. For these differences we accounted with logistic regression models and according matching strategies. Fourth, our study is a post-hoc analysis of an observational population-based design. Hence, a causal relationship cannot be made and mechanistic modelling is necessary. Furthermore, we are not able to rule out the possibility of unrecognized confounders. Preterm birth is associated with several postnatal complications that might impact neurologic development 29 . Potential factors such as antibiotic use, gut dysbiosis, nutrition, use of several drugs or unrecognized systemic reactions might have an impact on neurodevelopmental outcome which we cannot adjust  Table 3. Outcome characteristics of matched VLBWI at 6-year follow-up. Outcome of matched cohort for NEC or SIP. Mahalanobis distance matching criteria were GA, ICH, PVL, European origin, BPD, female gender, antenatal administration of steroids, and maternal education; p-values are derived from T-Test (IQ), Pearson's Chi-square test or Mann-Whitney U-test; the type I error level was set to 0.05; data are given as mean (SD) or n (%); significant findings (p < 0.05) are given in bold. www.nature.com/scientificreports www.nature.com/scientificreports/ for. Additionally, we were not able to control our analyses for short bowel syndrome as known risk factor for neurological impairments 18 as we did not record this disease in our follow-up examinations. To account for a variety of known confounders we used Mahalanobis distance matching as it is proved to be a valid approach for adjustments for multiple outcomes 30 . The matched cohort showed a homogenous distribution of factors for NEC which suggest that the model fits adequately. In the SIP group, gestational age and birth weight were significantly lower than in the control group. However, both groups do not differ concerning outcome characteristics, strengthening our conclusion that SIP even at lower GA has no impact on intelligence quotient.
In conclusion, our results reflect that the different aetiology and inflammatory extent of NEC and SIP may lead to disparate neurodevelopment trajectories. Hence our data suggest a potential role of an early gut-brain axis distortion in infants with NEC.
Future longitudinal studies, specifically in cohorts with interventions on the preterm infants microbiome such as PRIMAL 31 , along with detailed mechanistic models are needed to disentangle the impact of gut dysbiosis and sustained inflammation on adverse neurodevelopmental outcome after prematurity. Cohort and definitions. The German Neonatal Network (GNN) is an ongoing multicentre population-based cohort study enrolling VLBWI with <1500 g birth weight in Germany. From 2009 until 2014, 43 tertiary German tertiary level NICUs contributed to the GNN (www.vlbw.de) 32 . Data were collected prospectively by neonatologists or trained study personal. Infants <1500 g birth weight and with a gestational age below 37 weeks were enrolled. A predefined clinical data set including antenatal and postnatal treatment and outcome data was recorded prospectively. The proper assessment of clinical data was ensured by yearly on-site-monitoring by a study nurse or paediatrician experienced in neonatology. NEC requiring surgery is defined as clinical NEC classified as Bell Stage II or Bell Stage III with the need for laparotomy with or without resection of necrotic gut, and the macroscopic diagnosis of NEC 33 . Clinical NEC without surgical treatment was excluded from our analysis. SIP diagnosis was defined as the occurrence of spontaneous intestinal perforation with the need for laparotomy and the macroscopic diagnosis of isolated SIP as described by the attending surgeon. Small for gestational age (SGA) was defined as birth weight <10 th percentile according to population based birth weight reference values 34 .
Six-year follow-up. At the age of 5-6 years, VLBW children were invited for standardized follow-up examination by a study team consisting of a physician and nurses. During the invitation procedure, the study team contacted the clinic of birth for possible follow-up examination dates. The contact database was randomly searched for possible candidates with focus set on infants born <28 weeks gestational age, but infants born >28 weeks were not necessarily excluded. Infants who were born in the clinic of interest, contactable via telephone or postal letter and who could present at the follow-up examination appointment received an invitation letter.

Statistical analysis. Cohort characteristics.
To describe the characteristics of the whole cohort of VLBWI, we present differences of infants with NEC, SIP or without one of these complications. Data are presented as numbers, frequencies and 95% confidence intervals (CI) of column percentages.
Neurological outcome analysis. For neurological outcome analysis at 6 years of age, we included infants who were capable to perform a Wechsler preschool and primary scale of intelligence test (WPSSI-III, German edition). Logistic regression analysis and Mahalanobis distance matching were used to calculate a potential impact of NEC or SIP on outcome at six years with simultaneous controlling for several potential confounding factors. Mahalanobis' method is expected to be very successful in reducing bias in multivariate matching 31 and was used to compare NEC or SIP affected individuals with unaffected controls with similar risk profile.
Logistic regression analysis. Odds ratio and corresponding 95% confidence interval deriving from a logistic regression model was calculated to characterize an association of intelligence quotient <85 at 6 years of age and CP with neonatal complications as potential confounders: birth weight, gestational age (GA), female gender, NEC, SIP, intracerebral haemorrhage (ICH) grade ≥3 and periventricular leukomalacia (PVL), neurosurgery for ventriculoperitoneal shunting (post-haemorrhagic hydrocephalus), European ethnicity, bronchopulmonary dysplasia (BPD), born small for gestational age (SGA), surfactant application, and maternal education level. We calculated the relative risks (RR) derived from the logistic regression by using the proposed method from Zhang et al. 16 .
For multiple testing, p-values were adjusted using Bonferroni-Holm method and regarded significant when p < 0.05. Infants who were not capable for WPSSI testing are analysed separately using descriptive statistics. (2020) 10:2373 | https://doi.org/10.1038/s41598-020-58761-6 www.nature.com/scientificreports www.nature.com/scientificreports/ Mahalanobis distance matching. To analyse the effects on intelligence quotient and motor outcome, we matched the participants into two groups for each complication: NEC positive and NEC negative or SIP positive and SIP negative clinical courses. We matched the groups via Mahalanobis distance multi-dimensional modelling 38 . Matching was based on the calculated Mahalanobis distance, including gestational age, ICH, PVL, European ethnicity, BPD, female gender, antenatal administration of steroids, and maternal education for NEC and SIP analysis. For each index case with NEC or SIP, matches were chosen by the best fitting non-affected nearest partner using calculated Chi-squares of two distances.
All statistical analyses were performed with SPSS 22.0 software (IBM SPSS Statistics for Windows, Version 22.0. Munich, Germany). Graphics were created using GraphPad Prism Version 7.00 for Mac (GraphPad Software, La Jolla California USA, www.graphpad.com). ethics approval and consent to participate. Written informed consent was obtained from parents on behalf of the infants enrolled in our study. The study parts were approved by the local committee on research in human subjects of the University of Lübeck (08-022; 03.12.2010) and the local ethical committees at the other study centres.
Specifically consent for publication. Consent was given by the parents or legal guardians.

Data availability
The datasets generated and analyzed during the current study are not publicly available but can be reviewed on reasonable request.